1. Field of the Invention
The present invention relates to an inspecting apparatus utilizing a pattern recognition technique and, more particularly, to an apparatus suitably applied to enlarging or reducing each pattern size in binary image information.
2. Related Background Art
In a semiconductor fabrication apparatus, a pattern to be projected onto a semiconductor wafer is formed beforehand on a reticle. In order to inspect whether the pattern on the reticle has a predetermined size or whether scratches and unnecessary indentations are present in the pattern, the pattern on the reticle is picked up by an image pick-up device, and signal levels of pixels constituting a planar pattern are converted into binary data signals. The detected binary data is compared with reference data representing a reference pattern.
When the reticle pattern is inspected according to the above technique and the reticle pattern size is slightly different from the predetermined reference pattern, a defect present in the reticle pattern often cannot be properly be detected upon comparison between the detected binary data and the reference data, and the entire pattern is often detected as defective. In order to eliminate this drawback, the reference pattern must be enlarged or reduced, so as to be matched with the pattern to be inspected.
When a reticle pattern is designed in practice, individual pattern sizes are selected on the basis of a common reference pattern so as to comply with the fabrication steps of semiconductors to be fabricated. In this manner, a pattern having an optimal size must be used to form an identical pattern on a reticle or a mask. In this case, the pattern size must be controlled to derive a size of a pattern to be formed on the reticle or mask from the reference design pattern.
For this purpose, a typical conventional pattern processor is proposed wherein positional image information is generally produced by a computer and stored in an external memory, and the stored data is read out at the time of creation of a desired reticle pattern, thereby obtaining an image signal for forming the reticle pattern.
In order to enlarge or reduce the size of the pattern according to the above method the computer must perform arithmetic operations or binary image data for all patterns according to an enlargement or reduction ratio, thus undesirably requiring long processing time.